RTX 3070 FE low DX 12 performance

[Jice]

I just upgraded from a rx5700 to a RTX 3070 and ran pass mark 3D bencmark, yet my new card scored 10 FPS lower on average than my old card but only on DX 12. Should I be concerned since DX 12 is what most new games run on?

[David (PassMark)]

Did anything else change. e.g. monitor resolution?
What were the two DX12 scores?

[Jice]

Did anything else change. e.g. monitor resolution?
What were the two DX12 scores?

My RTX 3070 got 65-70 FPS 1080p
My rx 5700 got 80-85 1080p

the rtx did much better in ever other category though except 2d

[David (PassMark)]

85fps for the DirectX12 test for a RX 5700 is extremely high.
Nearly off the chart in fact.

Average is 55fps. Median is 35fps.




Where as your score for the 3070 is exactly in line with expectations.

This doesn't full explain things, but should mean it isn't a big concern.

[ColinB]

Those numbers look low to me:
I have the Gigabyte RTX3070 non-OC. I've pushed its memory and core clock up (basically to the OC variant level), and I've tuned my system RAM as best I can. Everything is air-cooled, so I'm not trying to go past the systems 'normal' capabilities.

That gives me a Passmark 3D Graphics score of 26638, and on DX12, a score of 163. I have no idea if that DX12 score is in 'fps' or not.
The monitor is 4k, running at 160fps.

[David (PassMark)]

Yes, the numbers for the DX12 test are frame per sec (fps). But in the case where the user can't get to the target 4K resolution (e.g. their monitor forces them to 800x600 pixels) then we scale the fps result down to compensate for the lack of resolution.

> My RTX 3070 got 65-70 FPS 1080p [in DX12]

> on DX12, a score of 163 [at 4K]

Those two statements seem to contradict each other. You get 65fps at 1080p but 163fps at 4K.

[Jice]

Yes, the numbers for the DX12 test are frame per sec (fps). But in the case where the user can't get to the target 4K resolution (e.g. their monitor forces them to 800x600 pixels) then we scale the fps result down to compensate for the lack of resolution.

> My RTX 3070 got 65-70 FPS 1080p [in DX12]

> on DX12, a score of 163 [at 4K]

Those two statements seem to contradict each other. You get 65fps at 1080p but 163fps at 4K.

This is the benchmarks, I only included the DX12 portion because the other ones benchmark higher than my rx 5700. Please any advice will help.
I have an Ryzen 7 3700X
16GB DDR4 XMP
Latest Drivers

[David (PassMark)]

As mentioned the RTX3070 appears to be scoring exactly on average for the DX12 test. So there nothing to investigate here and no advice to give (except to say, the result might be slightly higher at 4K compared to 180p as we won't scale the numbers down).

So the real question is why was your RX5700 score above average. Maybe overclocked, or maybe a fast CPU (as compared to the typical RX5700 user), maybe device driver settings. We don't know the exact answer and it is hard to investigate remotely.

[ColinB]

Yes, the numbers for the DX12 test are frame per sec (fps). But in the case where the user can't get to the target 4K resolution (e.g. their monitor forces them to 800x600 pixels) then we scale the fps result down to compensate for the lack of resolution.

> My RTX 3070 got 65-70 FPS 1080p [in DX12]

> on DX12, a score of 163 [at 4K]

Those two statements seem to contradict each other. You get 65fps at 1080p but 163fps at 4K.

Yes, those statements DO contradict each other.
The first is by the OP (with the problem), the second is mine. That's why I commented the way I did.

I use an LG 27GN950B 4k monitor with 10 bit (8-bit + frc) colour and Freesync + G-Sync.

I get a 3D Graphics score (in Performance Test 10.0) of about 26285, and a DX12 score of 162 using a Gigabyte Eagle RTX3070

When I turn off all my tuning enhancements (CPU/IF/Mem back to 3600, memory completely at XFR2 defaults), I get 25925. That's only -1.4% down (hardly worth the effort, for framerates!)
When I turn off all performance enhancements in NVidia's Control Panel (eg going from 'Performance' to 'Quality' throughout), the result is 24986, or about -4%. The DX12 number is 146.

As you can see from the AIDA64 'maximum' columns, the system is not at all stretched.
The GPU operates at a maximum of 219w (against another run: giving 24901, DX12 146.
This is Radeon RX6900XT level performance, according to Passmark.
I'm using a lower tier RTX3070 (220w-240w) at the lowest end (220w). I don't think there are any 3070s lower than that.

I can only think of one bit of technology that might be impacting this - compression over DisplayPort, providing higher framerates, or maybe G-Sync (which is a bit of a mystery to me...).

Regards,









Regards,
Colin

[ColinB]

Further to my tests yesterday, here are some DX12 results from changing refresh rates on the 4K monitor, on its DisplayPort connection: the files all start with DX12 (followed by the refresh rate).
After that, I switched to HDMI at 60 Hz. This compared exactly to DisplayPort.
Then I went for 25Hz over HDMI: this lowered the DX12 framerate to 74.

Conclusion: the DX12 test seems to be 'refresh-rate invariant' (good) until it drops to 25Hz (or similar). I ran the 25Hz test twice, just to check - it was identical at 74.

I can't get scores below 74 in DX12 on a lower-tier RTX3070 on its lowest power setting, for any refresh rate.
A score of 60, on a RTX3070 is not normal in 4K.








3 additional captures will be put in the next post.

[ColinB]

Here are the 3 next refresh rate values: 60 Hz in DisplayPort, 60 Hz in HDMI, 25 Hz in HDMI.








This is my last post on the subject.
Regards,
Colin